Skip to content

Latest commit

 

History

History
212 lines (148 loc) · 7.93 KB

tegraK1.md

File metadata and controls

212 lines (148 loc) · 7.93 KB
Raw

tegraK1

Install

Cuda

Dev Utils

gcc 7.x

deb http://ppa.launchpad.net/jonathonf/gcc-defaults-7.1/ubuntu trusty main # deb-src http://ppa.launchpad.net/jonathonf/gcc-defaults-7.1/ubuntu trusty main

apt update & apt install gcc-7 g++-7

need to manually chnage sym-linsk from gcc-4.8/g++-4.8 to gcc-7/g++-7

cmake

3.12.x Just install from sources: ./bootstrap make make install

Pocl

OpenCL software implementation buil info

software

build: cmake -DTHREADS_HAVE_PTHREAD_ARG=ON -DLCC_HOST_CPU=cortex-a15 -DENABLE_FP64=ON ..

CUDA

to build POCL with CUDA required LLVM 4.0+, while default in Ubuntu 14.04LTS is 3.8.

-DENABLE_CUDA=ON -DENABLE_FP64=OFF also need to disable shuffle_double in /test/kernel/CmakeLists.txt

  • line 288 kernel\test_shuffle_double
  • line 313 kernel\test_-shuffle_double

in case of error in CUDA build cumment cuStreamWaitValue32 related code and change event->command.run.kernel to event->command->command.run.kernel type.

pocl/lib/CL/devices/cuda/pocl-cuda.c:986:5: warning: format ‘%lu’ expects argument of type ‘long unsigned int’, but argument 4 has type ‘size_t’ [-Wformat=]
     POCL_ABORT ("[CUDA] Total constant buffer size %u exceeds %lu allocated\n",
     ^
/mnt/64GB/home/az/sdk/src/pocl/lib/CL/devices/cuda/pocl-cuda.c: In function ‘pocl_cuda_submit_node’:
/mnt/64GB/home/az/sdk/src/pocl/lib/CL/devices/cuda/pocl-cuda.c:1089:7: error: implicit declaration of function ‘cuStreamWaitValue32’ [-Werror=implicit-function-declaration]
       result = cuStreamWaitValue32 (stream, dev_ext_event_flag, 1,
       ^
/mnt/64GB/home/az/sdk/src/pocl/lib/CL/devices/cuda/pocl-cuda.c: In function ‘pocl_cuda_finalize_command’:
/mnt/64GB/home/az/sdk/src/pocl/lib/CL/devices/cuda/pocl-cuda.c:1343:40: error: request for member ‘run’ in something not a structure or union
       cl_kernel kernel = event->command.run.kernel;
                                        ^
/mnt/64GB/home/az/sdk/src/pocl/lib/CL/devices/cuda/pocl-cuda.c:1344:48: error: request for member ‘run’ in something not a structure or union
       pocl_argument *arguments = event->command.run.arguments;

in case of error of c99 related just move size_t i; dectaration from the loop

pocl/lib/CL/devices/cpuinfo.c:362:9: error: ‘for’ loop initial declarations are only allowed in C99 mode
         for (size_t i = 0; i < sizeof (vendor_list) / sizeof (vendor_list[0]); ++i)
         ^
/mnt/64GB/home/az/sdk/src/pocl/lib/CL/devices/cpuinfo.c:362:9: note: use option -std=c99 or -std=gnu99 to compile your code
/mnt/64GB/home/az/sdk/src/pocl/lib/CL/devices/cpuinfo.c:411:7: error: ‘for’ loop initial declarations are only allowed in C99 mode
       for (size_t i = 0; i < part_count; ++i)
       ^

need to build without -jX (i.e. single-threaded)

LLVM 4.0+

get llvm code

git clone http://llvm.org/git/llvm.git & git checkout release_70

cd llvm/tools & git clone http://llvm.org/git/clang.git & git checkout release_70 & cd ../..

cd llvm/projects & git clone http://llvm.org/git/compiler-rt.git & git checkout release_70 & cd ../..

cd llvm/projects & git clone http://llvm.org/git/test-suite.git & git checkout release_70 & cd ../..

mkdir -p build & cd build

config: cmake -D LLVM_TARGETS_TO_BUILD="ARM;NVPTX" -DCMAKE_BUILD_TYPE=Release -D CMAKE_C_FLAGS=-mcpu=cortex-a15 -D CMAKE_CXX_FLAGS=-mcpu=cortex-a15 ../llvm

glew

glew used by pocl to implement CL to GL. need to be installed. beete from git clone https://github.com/nigels-com/glew.git glew - since trysty glew support only GL 4.4, while tegra support 4.5 cd glew/auto & make & cd .. & make install

system updates

system headers from arm-xx directory should be sym-linked to /usr/include - probably due to Linux4Tegra 23/24 wasn't originally for K1

ln -s /usr/include/arm-linux-gnueabihf/sys /usr/include/sys
ln -s /usr/include/arm-linux-gnueabihf/asm/ /usr/include/asm
ln -s /usr/include/arm-linux-gnueabihf/bits/ /usr/include/bits
ln -s /usr/include/arm-linux-gnueabihf/c++/ /usr/include/c++
ln -s /usr/include/arm-linux-gnueabihf/gnu/ /usr/include/gnu
ln -s /usr/include/arm-linux-gnueabihf/hwloc/ /usr/include/hwloc
ln -s /usr/include/arm-linux-gnueabihf/python2.7/ /usr/include/python2.7
ln -s /usr/include/arm-linux-gnueabihf/a.out.h /usr/include/a.out.h
ln -s /usr/include/arm-linux-gnueabihf/expat_config.h /usr/include/expat_config.h
ln -s /usr/include/arm-linux-gnueabihf/ffi.h /usr/include/ffi.h
ln -s /usr/include/arm-linux-gnueabihf/ffitarget.h /usr/include/ffitarget.h
ln -s /usr/include/arm-linux-gnueabihf/fpu_control.h /usr/include/fpu_control.h
ln -s /usr/include/arm-linux-gnueabihf/ieee754.h /usr/include/ieee754.h
ln -s /usr/include/arm-linux-gnueabihf/zconf.h /usr/include/zconf.h

tegra K1 use hard-float only, so no gnu-stubs-soft.h - change /usr/inlude/gun/stubs.h to hf only

#if !defined __ARM_PCS_VFP
# define __ARM_PCS_VFP
# include <gnu/stubs-hard.h>
#endif
#if defined __ARM_PCS_VFP
# include <gnu/stubs-hard.h>
#endif

OpenCV

missed files

In case if opencv/CMakeDownlaods.txt contain errors, manual upload of missed files required:

scp -r build/opencv/downloads/xfeatures2d az@10.0.0.7:/home/az/sdk/src/opencv

build with CUDA

  • can be build with CUDA, which required gcc below 4.10
  • -D WITH_CUDA=ON and opencv_contrib required for cuda build with 4.x.x
  • need to clear install if use OpenCV after 2018/10
cmake -D CMAKE_BUILD_TYPE=RELEASE -D CMAKE_INSTALL_PREFIX=/usr/local -D ENABLE_NEON=ON -D ENABLE_VFPV3=ON -D BUILD_TESTS=ON -D BUILD_EXAMPLES=ON -D WITH_CUDA=ON -D OPENCV_EXTRA_MODULES_PATH=../../opencv_contrib/modules ../../opencv

Qt need to be build with Desktop OpenGL only - need evaluate/send bug report

may appear error like boostdesc_bgm.i not found - this file may not be downloaded during camke preparations - need to be placed in build/xfeatures2d/*

be sure to have -lOpenCL in /opencv-3.4.3/build/samples/opencl/CMakeFiles/example_opencl_opencl-opencv-interop.dir/link file with OpenCL build

Vulkan

get vulkan4tegra beta

- Extract the package to local dir
- Copy nvidia_icd.json into /etc/vulkan/icd.d/ *You may need to create this directory*
- Copy loader/libvulkan.so.1.0.3 to /usr/lib/arm-linux-gnueabihf/
- Create softlinks with below command:
$ ln -s /usr/lib/arm-linux-gnueabihf/libvulkan.so.1.0.3 /usr/lib/arm-linux-gnueabihf/libvulkan.so
$ ln -s /usr/lib/arm-linux-gnueabihf/libvulkan.so.1.0.3 /usr/lib/arm-linux-gnueabihf/libvulkan.so.

as a result - Vulkan API Version: 1.0.3

Assimp

cmake -DASSIMP_BUILD_TESTS=OFF ..

Qt

./configure --release -opengl desktop -opensource -confirm-license -prefix /usr/local -no-eglfs

Qt 5.12 can't complile with EGLFS

Other tools

x11vnc -forever to avoid exist after every session

Hardware

setup Tegra4Linux

  • tegrastats

RAM 714/1876MB (lfb 139x4MB) SWAP 0/2048MB (cached 0MB) cpu [25%,7%,1%,1%]@2218 EMC 3%@924 AVP 0%@81 VDE 120 GR3D 0%@252 EDP limit 0

where GR3D - GPU freq in MHz, cpu is CPU freq in MHz, EMC is memory

  • get info
# cat /sys/kernel/debug/clock/gbus/possible_rates
72000 108000 180000 252000 324000 396000 468000 540000 612000 648000 684000 708000 756000 804000 852000 (kHz)
  • set maximum GPU freq
echo 852000000 > /sys/kernel/debug/clock/override.gbus/rate
 echo 1 > /sys/kernel/debug/clock/override.gbus/state

Camera

OmniVision OV5693 - front and back driver for intel Atom - use Atom ISP